Contrib Mineral Petrol (1996) 126: 2537

C Springer-Verlag 1996

L.Ya. Aranovich R.G. Berman

Optimized standard state and solution properties of minerals

II. Comparisons, predictions, and applications

Received: 14 September 1994

Accepted: 20 March 1996

Abstract Internally consistent thermodynamic data for endmembers and solid solutions derived in Part I of this work are used to predict phase relationships and compositions of solid solutions in good agreement with direct experimental observations that were not used in the calibration. This external consistency with a large body of experimental observations increases confidence in applications to both petrogenetic grid and thermobarometric calculations. The predicted position of the univariant FMAS equilibrium Gt Cd Opx Si Qz occurs between 7.8 kbar 700 C and 10.6 kbar 1100 C, in excellent agreement with available phase equilibrium reversals, and with thermobarometric results for both higher-P Opx Si Qz bearing and lower-P Cd bearing granulites. In these assemblages, garnet composition is an excellent geobarometer and Al2O3 content of Opx is an excellent geothermometer, almost independent of other compositional variables. Comparison of temperature estimates from different exchange and net-transfer reactions for a number of samples representative of highgrade terranes demonstrates the ability of carefully chosen portions of Fe Mg minerals to preserve information regarding a high temperature steady state during their evolution. The equilibrium Fs Ok Alm, based on the Al2O3 content of Opx, offers the most robust thermometry for the Opx Gt assemblage because of its relative insensitivity to late Fe Mg exchange. Applications of this thermometer indicate that in many samples the Al2O3 content of Opx yields very similar temperatures to Gt Cd Fe Mg exchange. For some high temperature samples, these temperatures are up to 150 C higher than calculated with the Gt Opx Fe Mg exchange thermometer.
L.Ya. Aranovich Institute of Experimental Mineralogy, Chernogolovka, Russia R.G. Berman (u) Geological Survey of Canada, 601 Booth Street, Ottawa, Ontario, Canada K1A 0E8 Editorial responsibility: K. Hodges

Introduction
Since the pioneering works of Karpov et al. (1976) and Helgeson et al. (1978), there has been a great deal of focus on the use of internally consistent thermodynamic data for quantitative petrologic calculations (e.g. Powell and Holland 1985, 1988; Holland and Powell 1990; Berman 1988, 1991). While experimental phase equilibrium studies have provided the foundations for quantitative understanding of igneous and metamorphic rocks, thermodynamic data provide the means to extrapolate from the laboratory to nature. This extrapolation is particularly important and challenging for calculation of phase relationships in complex chemical systems (e.g. Aranovich 1983, 1991) which afford closer approximations of natural systems than the simplified systems most amenable to producing interpretable experimental data. Extrapolation is also important for geothermobarometry calculations in which the deviations of mineral compositions from those produced in experimental studies are significant, and the P T range of experimental accessibility is often well removed from the physical conditions of rock equilibration. The potential drawback of the internally consistent approach to quantitative calculations is that the data may not represent all experimental results sufficiently well to meet the most stringent extrapolation requirements. It is therefore important to demonstrate that no loss of accuracy is introduced compared to the use, for example, of a geothermometer based solely on the results of one experimental study. We show in the companion paper (Berman and Aranovich 1996; referred to as Part I) that the combination of derived standard state and mixing properties represents most experimental observations within the uncertainty of the measurements; we suggest that our calibration based on data for many different equilibria results in an increase in overall accuracy over that obtainable from more limited data sets that, in most cases, cannot separate between the standard and mixing energy contributions. This accuracy can also not be obtained from sets of endmember ther-

26

modynamic data that are not linked with mixing properties through appropriate analysis of experimental data (Berman 1988; Holland and Powell 1990). In this paper we explore some of the geological implications of the optimized thermodynamic data for olivine, orthopyroxene, garnet, cordierite, and ilmenite. We first present calculated phase relationships in complex chemical systems that serve as a basis for consideration of many features in high-grade metamorphic assemblages of pelitic to intermediate composition, and compare these calculations with natural observations. Comparison with available experimental data will also serve as additional tests of the thermodynamic data derived in Part I. We then discuss the implications of the thermodynamic data for thermobarometric calculations for the same assemblages.

Calculated phase relations

All computations were performed on a PC-486 with the Gibbs free energy minimization software, THERIAK, described by de Capitani and Brown (1987). Thermodynamic data are those described in detail in Part I combined with internally consistent data for other minerals and fluids given by Berman (1988).
Fig. 1 Comparison of available experimental data with P XFe diagram for the divariant assemblage Gt Ilm Rt Si Qz in the FMAST system, computed with Part I thermodynamic data. Symbols show nominal experimental data, with diamonds and triangles showing Gt and Il compositions, respectively. Lines connected to symbols show effect of +0.01 XFe compositional uncertainties as well as direction from which final compositions were approached

Phase relations involving ilmentite Comparisons with experimental data discussed in Part I indicate that our derived properties for the ilmenite solid solution with low hematite component ( 0.03 Xmax) Hm yield reasonable representation of the energetics on the Ilm-Gk (see abbreviations in Table 1, part I) binary. Predicted phase relationships involving ilmenite provide an opportunity to further test our calibration, in addition to exploring some implications for geothermobarometry. Koziol and Bohlen (1992) examined displacement of the GRAIL equilibrium: 3 Ilm Si 2 Qz Alm 3 Rt (a) Although Koziol and Bohlen (1992) did not attempt to reverse the Fe Mg partitioning between garnet and ilmenite, the consistency of their experimental observations (Fig. 1) suggests a close approximation to equilibrium. Koziol and Bohlen (1992) note that their data yield KD 4.9, while Green and Sobolev (1975) measured KD values around 4.0 in synthesis experiments on pyrolite and olivine-basanite compositions. Our calculations are in reasonable agreement with both these studies, with KD increasing from 3.6 to 5.2 as XFe increases from 0 to 1. A similar compositional dependence of KD is computed for Fe Mg exchange between Opx and Il. At 1000 C, KD increases from 8 to 13 going from XFe 0 to 1. At higher temperatures, this compositional dependence is reduced as the ilmenite solid solution becomes closer to ideal. Our calculated KD values reproduce within uncertainties the 8001100 C measurements of Hayob et al. (1993) which were utilized in Part I, as well as the majority of the 8001300 C data of Bishop (1980) which were not included in the Part I calibration (Fig. 2). Although only two bulk compositions were studied, the latter data clearly show the large compositional dependence of KD, which is important to account for in accurate thermodynamic calculations. One half-bracket of Bishops at 1000 C indicates much higher KD than our predictions and than the data of Hayob et al. This half-bracket appears to represent a large overstepping of the equilibrium composition produced by the use of highly metastable

through addition of MgO, and concluded that Fe Mg mixing in garnet is very near ideal. They report incorporation of up to approximately 2 weight % TiO2 in run product garnet, while Bohlen et al. (1983), in studying equilibrium (a) in the Mg-Free system, considered TiO2 in garnet analyses to be due to dispersed Rt inclusions. Because of the ambiguity between these observations, we did not use the data of Koziol and Bohlen (1992) in the calibration described in Part I. Predicted phase relations show excellent agreement with the experimental observations (Fig. 1). The garnet side of the divariant loop lies within the uncertainties of, but at slightly less Fe-rich compositions than those determined by Koziol and Bohlen, consistent with the calculations which do not account for solution of TiO2 in garnet.

27

Fig. 2 Variation of KD with XFs at 13 kbar for the equilibrium Gk Fs En Ilm. Symbols as in Fig. 1 Fig. 3 Mg-rich portion of computed Roseboom diagram for Ol Opx at P 2 kbar, and 400-800 C. Inset shows isotherms over entire compositional range. Compositions of natural coexisting Ol Opx (filled circles) taken from compilation by Sack and Ghiorso (1989). Note that for XFe 0.10, Part 1 systematics predict XOpx XOl at temperatures below 800 C Fe Fe

starting materials consisting of a mixture of endmember geikilite ferrosilite. The above comparisons indicate that ilmenite properties derived in Part I afford very reasonable approximation of phase equilibrium data involving low Fe3 ilmenite. Fe Mg exchange thermometers have been proposed for the pairs Ol-Il (Andersen et al. 1991) and OpxIl (Bishop 1980). The partitioning of Fe Mg between Il and Cd is even more extreme, presenting an attractive possibility for geothermometry, albeit limited by the low concentration of Mg in ilmenite at low temperature. Our calculations again show that deviations from ideality in these phases lead to a strong compositional dependence with KD increasing from 4575 over the range XFe 0 to 1 at 600 C and from 18 to 30 over the same compositional range at 800 C. Stability of orthopyroxene olivine ( quartz)

Figure 3 shows calculated Fe Mg exchange isotherms for coexisting olivine-orthopyroxene over the temperature range 400800 C. In accord with the experimental data on which they are in large part based (von Seckendorff and ONeill 1993), our calculations indicate (1) a very small temperature dependence of KD, which reflects the lack of suitability of this exchange couple as a geothermometer, and (2) that Ol is more Fe-rich than Opx at all temperatures above 700 C. Crossover of the computed isotherms with the diagonal line representing equal partitioning of Fe and Mg between Ol and Opx, such that Ol becomes less Fe-rich than Opx, has been inferred from some natural observations (see references listed in Sack and Ghiorso 1989) and from older experi-

mental studies (Medaris 1969; Matsui and Nishizawa 1974; Fonarev 1987; Koch-Mueller et al. 1992). Reasons for the apparent disagreement with these experiments have been discussed in detail by von Seckendorff and ONeill (1993). We can add to their arguments that only two experimental points with XOl XOpx (XOl Fe Fe Fe 0.11, 900 C; XOl 0.11, 1000 C, Koch-Muller et al. Fe 1992) are inconsistent with the calculated isotherms (Fig. 3) in terms of direction of approach to equilibrium (see discussion in Part I). These inconsistencies can be attributed to experimental difficulties such as X-ray determination of final compositions, and possible lack of equilibrium due to extreme sluggishness of the reactions with starting Mg-rich Opx. Apparent disagreement with the natural observations deserves more detailed consideration. According to our calculations there is a point, at approximately XOpx 0.28 (see inset in Fig. 3), of intersecFe tion of the exchange isotherms, which divides them into two parts with the opposite temperature dependence (i.e. with the opposite sign of the integral enthalpy change of the exchange reaction). Such a crossover point was not observed in the experiments of von Seckendorff and ONeill (1993), but it is not precluded by them either (cf. runs 41, 42, 13 1, 13 2, 1 1, 33, 38, and 34 in their Table 2). Low-temperature extrapolation of our fit for these extremely Mg-rich compositions produces a crossover with the diagonal (Fig. 3) at temperatures below ca.

28

700 C. For natural samples with XOl less than or equal to Fe XOpx (compositions taken from Sack and Ghiorso 1989, Fe their Table A1), our calculations predict equilibration temperatures in the range 400750 C, in broad agreement with the geological environments in which they formed (serpentinites and tectonically emplaced peridotites). These samples may well have undergone a process of low temperature reequilibration, rather than representing complete disequilibrium due to different reaction rates of Ol and Opx with a hypothetical Mg-reservoir as suggested by von Seckendorff and ONeill (1993). The olivine-orthopyroxene-quartz assemblage was one of the first quantitatively calibrated geobarometers (Bohlen et al. 1980; Bohlen and Boettcher 1981), and our calculations (Fig. 4) confirm its validity as a geobarometer. It should be noted that, although our predicted olivine composition in equilibrium with orthopyroxene quartz at any specific temperature is slightly less Fe-rich than determined in these studies (see Part I), the position of the Fe (Fe Mg) isopleths on a P T diagram (Fig. 4) coincides well with those determined experimentally by Bohlen and Boettcher (1981). In spite of the fact that the Opx isopleths have a very shallow dP dT slope, suggesting a minimal temperature dependence of pressure readings for this geobarometer, there is an indirect influence of temperature. For example, a rock with bulk Fe (Fe Mg) 0.8 equilibrated at 800 C and 4 kbar should contain XOpx 0.78 and Fe XOl 0.88 (Fig. 4, see also Fig. 7, Part I). Low-temperaFe

ture Fe Mg resetting down to 600 C (without operation of the net-transfer reactions) would change coexisting compositions to XOpx 0.76 and XOl 0.90, leading to a Fe Fe pressure determination of less than 2 kbar (Fig. 4). This implies that pressure estimates from the Ol Opx Qz assemblage in rocks subjected to slow cooling may correspond to a minimum rather than peak metamorphic pressure. Stability of orthopyroxene magnetite quartz The equilibrium: 3 Fs 1 2 O2 Mt 3 Qz (b)

limits the maximum Fe-content of Opx at oxygen fugacities above that of the QFM buffer, and is important as a oxybarometer for high-grade metamorphic rocks of a variety of bulk compositions (Fonarev 1987). Experimental observations on the composition of Opx equilibrated with Mt and Qz at temperatures between 700 and 850 C, pressures of 1 and 5 kbar, and fO2 buffered by Ni NiO were reported by Fonarev et al. (1976) and Fonarev (1987). These data are important for independent assessment of the mixing properties of Opx on the En-Fs join, as well as the standard state properties of ferrosilite. They were not incorporated into the analysis of Part I because of ambiguities in the experimental determination of Opx composition. Contradictory interpretations have been given to the data by Fonarev (1987) and Lindsley (1980); the former suggesting a positive dT dXFe slope of the phase boundary, and the latter a negative slope. Figure 5 shows T XFe sections at two pressures that were experimentally investigated, calculated for this assemblage with the thermodynamic parameters derived in Part I. Our systematics predict a very slight negative T XFe, a rather diplomatic compromise to Lindsleys (1980) and Fonarevs (Fonarev et al. 1976; Fonarev 1987) interpretations. More importantly, however, our calculations are consistent with the experimental obser-

Fig. 4 Comparison of computed isopleths of Opx (labelled at 5 mol% Fs intervals) with experimental observations by Bohlen and Boettcher (1981). Triangles show P T conditions, adjusted for experimental uncertainties, at which Opx of the corresponding Fs content was found stable. Squares show conditions where Opx broke down to the divariant assemblage Opx Ol Qz. Opposite ends of attached lines show nominal experimental P T conditions

Fig. 5 T XFe diagram for the assemblage Opx Mt Qz at oxygen fugacities buffered by Ni NiO and pressures of 1 and 5 kbar. Labelled curves calculated with thermodynamic data from Part I. Symbols as in Fig. 1, with data at 1 kbar (solid symbols) and 5 kbar (open symbols) of Fonarev et al. (1976, triangles) and Fonarev (1987, squares)

29

vations on compositional changes of Opx during the course of reaction (b), particularly if the compositions of Opx reported by Fonarev (1987), rather then those reported in the earlier publication (Fonarev et al. 1976), are taken. Garnet-orthopyroxene phase relations in CMAS Experimental observations on stability of pyropegrossular garnet relative to Opx An Sil in the quartzsaturated portion of the system CMAS (Hensen 1976; Perkins 1983) were not included in the thermodynamic data retrieval described in Part 1, and provide an independent test for the derived garnet solution properties as well as standard properties of the minerals involved. The position of the univariant curve together with computed XGt and XOpx 3, are shown in Fig. 6. Agreement Gr Al2O between our calculations and independent experimental results is reasonably good. Our calculations indicate 7% grossular component at 1300 C and 15.3 kbar (Fig. 6), in excellent agreement with experimental observations (67% grossular) at the same P T conditions (Perkins 1983), Our calculated curve is at slightly lower pressure than the lowest temperature brackets obtained by Perkins (1983), but is consistent with Hensens (1976) data. On the basis of their own thermodynamic analysis, the valid-

ity of Perkins brackets was questioned by Perkins (1983) and Wood and Holloway (1984). Some peculiarities of the univariant boundary are worth noticing here: the strong variation in the curvature of the univariant boundary, and the significantly lower solubility of grossular component on the sillimanite side of the curve. Both these features stem from the fact that the univariant reaction is the extremal type (Prigogine and Defay 1954; Korzhinskii 1973). At a certain bulk composition, reactions of this type become degenerate, involving a lesser number of phases than in the general case. For the reaction under consideration this composition is close to Py67Gr33, where Al2SiO5 is not needed to balance the reaction. Small deviation from this composition is caused by appreciable solubility of Al2O3 in Opx. Garnet-orthopyroxene-cordierite phase relationships in FMAS The P T diagram portraying phase relations between garnet, orthopyroxene and cordierite sillimanite, quartz (Fig. 7) has many important implications as well as representing an extremely sensitive test of thermodynamic properties derived in Part I. The principal feature of this diagram is the univariant curve separating Opx Si from Gt Cd assemblages in silica-saturated rocks. The lowest pressure of this boundary, computed with the data of Part I, is approximately 8 kbar, indicating the relatively high pressure origin of hypersthene-sillimanite-quartz gneisses described in many ancient granulite-facies terranes (Harley 1989 and references

Fig. 6 Comparison of experimental observations by Hensen (1976, diamonds) and Perkins (1983, squares) in the system CaO MgO Al2O3 SiO2 with univariant curve (bold) for the equilibrium Gt Qz Opx An Sil Ky, calculated with thermodynamic data of Part I. Numbers above and below the curve show XGr and XOk, respectively. Symbols as in Fig. 4. Pronounced curvature of the reaction is caused by compositional degeneracy in which the Ca: Mg ratio in Gt becomes identical to that of the assemblage Opx An at XGr 0.33)

Fig. 7 P T grid for the assemblage Gt Cd Opx Sil Qz in the system MgO FeOAl2O3 SiO2. Curves computed with Part I thermodynamic data (with fully hydrated cordierite) show univariant equilibrium (bold curve), as well as isopleths of 100 XGt (numFe bers inside boxes) and Al2O3 content of Opx (1 mol% intervals, ovals). P T positions of natural samples with the most Mg-rich Gt (Mg 51) in the assemblage Gt Opx Crd Qz and the most Fe-rich Gt (Mg 57) in the assemblage Gt Opx Sil Qz are plotted as filled square and circles, respectively. Solid and open diamonds show nominal experimental brackets (assuming 10% friction correction for talc pressure cells) for the univariant equilibrium from Bertrand et al. (1991)

30

therein). It also shows that formation of the Opx Si Qz assemblage is restricted to metapelite compositions with relatively high bulk Mg (Mg Fe) ratios ( 0.5). Although generally similar to the previously published grids (Aranovich and Podlesskii 1989; Hensen and Harley 1990), our calculations (Fig. 7) differ from the previous studies in some important details. First, we predict the transition at 12 kbar lower pressure, and correspondingly more Fe-rich garnet composition along the univariant curve. According to our calculations the lower limit for Mg of garnet equilibrated with Opx Si Qz is about 50, compared to 70 calculated by Aranovich and Podlesskii (1989) and about 65 inferred by Hensen and Harley (1990). Support for the grid presented here comes from reports of garnet as low in Mg as 57 and 55 (circles in Fig. 7) in Opx Si Qz gneisses from the Aldan Shield (Perchuk et al. 1985; Aranovich and Podlesskii 1989) and Varpaisjarvi, Central Finland (Holtta et al., work in preparation), respectively. In addition, the predicted univariant boundary is in excellent agreement with high-temperature reversals on this reaction obtained by Bertrand et al. (1991). Unfortunately, run products of this study were compositionally extremely heterogeneous, preventing meaningful comparisons with our predicted compositions. Our calculated compositions of Opx and Cd along the univariant curve do not reveal the maxima in Mg suggested by Hensen and Harley (1990). Rather they show a steady increase in Mg with increasing temperature. The computed curvature of the univariant boundary is also not as pronounced as that proposed by Hensen and Harley (1990). The extremal nature of the curve (a pressure maximum exists above 1100 C) is dictated in this case by the fact that the bulk Mg (Mg Fe) ratio of the low pressure assemblage Gt Cd becomes equal to that of Opx. Garnet Mg in the divariant assemblages Gt Cd Opx Qz, Gt Cd Sil Qz and Gt Opx Sil Qz is extremely sensitive to pressure variations (Fig. 7) and, therefore, can be used as a geobarometer independent of any other compositional parameters. The only cause for caution is in the pressure dependence of the isopleths on the activity of cordierite which itself depends on water activity. Our calculations show that the garnet isopleths, which correspond to the divariant equilibrium Gt Qz Opx Cd, can be a maximum of 2 kbars lower in pressure in the limiting case of anhydrous cordierite. Cordierite stability is known to be significantly influenced by the presence and composition of a fluid phase (e.g. Newton 1972; Aranovich and Podlesskii 1989). Both dry (Pfluid 0) and pure CO2 (P PCO2) conditions lead to an increased stability field of the Opx Si Qz assemblage at the expense of Cd Gt. It is important to note, however, that at higher pressure, where the solubility of H2O and CO2 molecules in the cordierite structural channels becomes almost identical, the difference between the location of the univariant curve becomes very small, as is well illustrated in the experimental observations by Bertrand et al. (1991).

Isopleths of Al2O3 in orthopyroxene have a negative slope on P T diagram (Fig. 7), which differs strongly from the previously published results (Aranovich and Podlesskii 1989; Hensen and Harley 1990). This difference is caused primarily by a significantly lower Al-content in Opx in FAS predicted by the present study in accord with new experimental observations by Aranovich and Berman (1995, 1996). It may have important implications for deciphering reaction textures in highgrade rocks: for assemblages lacking primary Opx, development of Cd Opx coronas around Gt should be indicative of decompression, while for those having matrix Opx it could be just as well related to cooling. Problems associated with applications of Al-in-Opx thermometry will be further discussed below under geothermobarometry. Cordierite has been reported in textural equilibrium with the assemblage garnet-orthopyroxene-quartz with garnet as magnesian as Mg 51 (square in Fig. 7; Sutam block of the Aldan Shield; Aranovich 1991) and Mg 54 (Enderby Land, Antarctica; Ellis et al. 1980). If fluid-absent conditions had dominated during granulite-facies metamorphism, the transformation from cordierite- to sillimanite-bearing garnet-orthopyroxene assemblages should have taken place at much more Ferich garnet compositions (Aranovich and Podlesskii 1989, their Fig. 5). This implies that at least in some areas a fluid phase was physically present during the high-grade metamorphic event, a conclusion reached also by Hensen and Harley (1990) on similar grounds.

Geothermobarometry implications
In this section we highlight a few implications associated with application of the systematics derived in Part I to deciphering the P T conditions of equilibration of highgrade metamorphic assemblages. These calculations have been made using the TWQ software of Berman (1991). Garnet-orthopyroxene geothermometer Geothermometry based on the exchange of Fe Mg between garnet and orthopyroxene, via the equilibrium: 3 Fs Py Alm 3 En (c)

has been experimentally investigated in several laboratories (Kawasaki and Matsui 1983; Harley 1984; Lee and Ganguly 1988; Eckert and Bohlen 1992). Our calculations summarized in Part I demonstrate that derived thermodynamic properties agree within uncertainties with almost all data from these four studies that involved crystalline starting materials. Because of the use in Part I of the entire combined experimental data set, our calibration for equilibrium (c) is significantly different from any based solely on one set of experiments. Termperatures computed for coexisting garnet-orthopyroxene pairs in a variety of geologic environments

31 Table 1 Temperatures computed for natural samples with the Fe Mg Opx Gt exchange equilibirum (c) Area Reference Aldan Shield Aranovich (1991) Furua Complex Coolen (1980) Asuanipi Complex Percival (1991) Oaxaca Mora and Valley (1985) Adirondacks Jaffe et al. (1978) Arendal, Norway Lamb et al. (1985) Bamble, Norway Harlov (1992) Adirondacks Jen and Kretz (1981) Antarctica Sandiford 1985) Otter Lake Perkins et al. (1982) Minto Block Begin and Pattison (1994) Nain Complex Berg (1977) 3 13 12 11 10 4 2 6 10 5 4 7 6 3 6 5 17 12 Average 860 797 809 752 763 765 665 750 699 714 680 708 671 742 691 713 811 735 Range 825880 652919 743919 643829 694829 700809 619711 720793 646746 653851 653707 630988 630760 666844 580800 654800 6311140 631785 XGt Fe 0.350.52 0.590.75 0.590.75 0.580.70 0.580.69 0.720.77 0.980.99 0.660.85 0.600.90 0.850.89 0.850.87 0.480.82 0.480.82 0.770.80 0.640.79 0.640.79 0.640.85 0.640.85 XGt Gr 0.010.02 0.160.17 0.160.17 0.030.04 0.030.04 0.070.19 0.21 0.090.19 0.020.20 0.170.22 0.180.22 0.020.20 0.020.20 0.180.21 0.030.08 0.030.08 0.010.05 0.010.05 XOpx Ok 0.050.11 0.020.03 0.020.03 0.050.07 0.050.07 0.010.03 0.0030.005 0.010.03 0.010.06 0.013 0.013 0.010.11 0.010.11 0.0080.009 0.020.06 0.030.06 0.030.06 0.030.06

are presented in Table 1. If compositional zoning in garnet and or orthopyroxene was reported in the original publications, temperature estimates have been calculated using compositions of the cores of the minerals. The results in Table 1 indicate that our systematics produce T estimates broadly consistent with the petrologic intuition mineral facies approach. The temperatures obtained for each complex generally fall within +75 C of the average value, indicating reasonable clustering of the Testimates but not without the effects of late stage reequilibration in some samples. We consider this an encouraging feature of the systematics given the wide range of Gt and Opx compositions covered by the data in Table 1. Especially notable are the temperature estimates for Adirondaks samples with extremely Fe-rich garnet and orthopyroxene (XFe 0.980.99, Jaffe et al. 1978), which are in excellent agreement with temperatures computed for more typical compositions found in the Adirondaks (Johnson and Essene 1982). In general, our computed temperatures average about 3050 C lower than those based on the calibration of Lee and Ganguly (1988). A central point of contention in recent applications of geothermobarometry concerns the ability of Fe Mg exchange thermometers to recover near-peak temperatures for relatively slow cooled rocks (e.g. Pattison and Newton 1989; Frost and Chacko 1989; Harley 1989; Fitzsimons and Harley 1994; Pattison and Begin 1994). Detailed compositional maps which show large, steep gradients in Fe (Fe Mg) at contacts between ferromagnesian minerals, and little to no gradients where garnet is

surrounded or in contact with quartz or feldspar (e.g. Pattison and Begin 1994; Florence and Spear 1991) leave little doubt that low temperature reequilibration via diffusive exchange is a common, if not unavoidable feature in slowly cooled metamorphic rocks. These observations are also supported by theoretical simulations of diffusion profiles in minerals (e.g. Chakraborty and Ganguly 1992). Considerable uncertainty remains, however, regarding the extent to which these low-temperature processes affect core compositions of low-diffusive minerals, in particular garnet, which is involved in the most effective geothermometers. As pointed out by Pattison and Begin (1994), kinetic modelling does not provide an unambiguous solution to this problem because cooling rates of the rocks are unknown and uncertainties of experimentally determined diffusion coefficients are large enough to permit contradictory results from diffusion profile simulations. Our calculations do not reveal any significant temperature underestimates suggested by some authors to be a typical feature of the Fe Mg exchange thermometers in high-grade rocks (e.g. Pattison and Newton 1989; Frost and Chacko 1989; Harley 1989), although a certain amount of lower temperature resetting is indicated for a few samples in Table 1 taken by themselves. For less ambiguous conclusions to be drawn, computed Fe Mg exchange temperatures must be compared with estimates based on a compositional parameter assumed to be more conservative with respect to low-temperature reequilibration. The Al2O3 content of Opx has been suggested by

32

Fig. 8 Difference between temperatures computed with Al2O3 content of Opx (squares TFe Al, triangles TMg Al) versus XOk in Opx (a), and TFe Mg (b). Zero line denotes ideal correspondence of the thermometers. Note the much closer correspondence to TFe Mg of TFe Al than TMg Al. c P T diagram showing linear dependence among the exchange equilibrium Py Fs En Alm and the two net transfer equilibria Fs Ok Alm and En Ok Py

many authors (Aranovich and Podlesskii 1989; Anovitz 1991; Fitzsimons and Harley 1994; Pattison and Begin 1994) to be less sensitive to low temperature reequilibration because it is governed by net transfer reactions that have much higher activation energies than the Fe Mg interdiffusion process. In order to evaluate these suggestions, we have computed temperatures for the samples in Table 1 not only on the basis of the Fe Mg exchange equilibrium (c), but also the net transfer equilibria: 3 Fs Al2O3 Alm 3 En Al2O3 Py (d) (e)

Equilibrium (e) has been used extensively as a geobarometer in ultramafic systems (e.g. Finnerty and Boyd 1984) and a geothermometer in more acid compositions (e.g. Pattison and Begin 1994; Fitzsimons and Harley 1994). To our knowledge equilibrium (d) has not been applied previously as a geothermometer. Here we refer to temperatures calculated with equilibria (c), (d) and (e) as TFe Mg, TFe Al, and TMg Al,respectively. For the samples in Table 1 we calculated temperatures using all three equilibria. The results of these calculations, expressed as a difference between temperature estimates from the net-transfer and exchange reactions are plotted as a function of XAl2O3 in Opx (Fig. 8a) and

TFe Mg (Fig. 8b). Observation of Fig. 8a shows that (TFe Al TFe Mg) is always less than (TMg Al TFe Mg), reflecting the fact that these three equilibria are not independent, with reaction (d) having a P T slope intermediate between those of (c) and (e) (Fig. 8c). In addition, the larger standard enthalpy change of equilibrium (d) ( 35.5 kJ mol) makes it more robust relative to equilibrium (e) ( 4.7 kJ mol). For these reasons, subsequent discussion will only consider the more reliable temperatures based on TFe Al and TFe Mg. In Figs. 8a and 8b, individual points are rather equally distributed around the zero base line that denotes perfect agreement between the two thermometers, suggesting no tendency of the exchange thermometer to underestimate temperature. About 70% of the squares plot within 75 C of the zero base line, demonstrating reasonably good agreement between the two thermometers. That reaction (d) appears to monitor somewhat higher T values than the exchange reaction (c) (Fig. 8b) reaffirms that the Fe Mg exchange freezes in at a lower temperature than Al net transfer. On the other hand, one may question whether equilibrium has been attained between Gt and Opx in this temperature range, particularly for the most Fe-rich samples with very low Al2O3 contents of Opx (see Table 1). TFe Mg values for these samples range from 770 to 640 C, in reasonable agreement with corresponding regional estimates (Johnson and Essene 1982) for the Adirondacks, while TFe Al is far too low for many samples to be explained with any equilibrium model. For the samples with relatively high TFe Mg, the net-transfer reaction (d) in many cases gives lower temperatures estimates than the exchange reaction (Fig. 8b). The ability of Fe Mg exchange thermometers to recover near-peak metamorphic temperatures in highgrade rocks can be further assessed by detailed compari-

33

Fig. 9 Thermobarometric results for sample LRD-72 (Berg 1977) computed with TWQ software and thermodynamic data derived in Part I

sons of the temperature readings of these two thermometers in individual rocks. We performed such tests on well documented samples of Gt Cd Opx Qz gneisses from the contact aureole of the Nain anorthosite massif (Berg 1977), as well as granulites of the Minto block, Northern Quebec (Begin and Pattison 1994) and Aldan shield, Eastern Siberia (Perchuk et al. 1985; Aranovich 1991). Most samples (e.g. Fig. 9) show good agreement between garnet-cordierite and garnet-orthopyroxene exchange temperatures (T2 and T3, respectively, of Table 2), with a general tendency (excluding Nain samples with zoned Gt) towards slightly lower temperatures (0 100 C) for the Gt-Opx thermometer. Gt-Cd exchange temperatures are, however, in excellent agreement with temperatures based on equilibrium (d). As the latter are much less sensitive than the Gt-Opx exchange equilibrium to late Fe Mg reequilibration, the lower Gt-Opx exchange temperatures may indicate somewhat higher Fe Mg diffusivity in Opx relative to Cd and Gt in this temperature range. Several samples of Bergs (1977) give unrealistically high temperatures with both thermometers when core compositions of the garnets were used in the calculations (Table 2). Berg reports that symplectic coronas of Opx Cd around compositionally zoned garnet are a

Table 2 Computed P T values for selected natural samplesa

Sample LRD-72 KI 3557 2-893 2-1833 2-1455b 2-1637b KI 3909 2-625 2-1726 RAW-437 2-1572b 2-1578b 2-1480b NK 420B NU-69 2-275 74-98a 2-1572r C 10Bc B 69Ec P 88c P 69c B 74B B 58 S 2d P 79d KU14316e Sut-2 Sut-61e Tok-18e Tok-18 1e

Ref Berg (1977)

XOk 0.039 0.042 0.27 0.056 0.054 0.054 0.048 0.039 0.034 0.026 0.047 0.032 0.039 0.044 0.043 0.042 0.047 0.044 0.033 0.027 0.040 0.022 0.58 0.046 0.770 0.731

XPy 0.348 0.140 0.188 0.191 0.335 0.206 0.163 0.165 0.172 0.168 0.292 0.244 0.285 0.224 0.213 0.231 0.146 0.230 0.187 0.188 0.225 0.206 0.346 0.320 0.344 0.299 0.546 0.465 0.643 0.551 0.551

P1 T1 6.4806 3.0870 4.2780 3.4887 7.2980 4.4935 3.8890 2.8800 3.1790 2.7735 6.3915 5.8850 6.3885 3.9810 3.2780 4.1810 2.7850 4.5845 5.6840 4.6745 7.3910 3.2650 5.3800 4.8770

T2 @ P 1 793 770 782 778 1170 920 855 690 710 645 990 985 1010 690 635 730 725 780 765 690 812 570 670 650

T3 @ P 1 805 875 820 785 900 830 860 750 745 730 910 860 890 755 690 735 780 810

P2 T4 6.4 797 3.5820 4.1800 4.3780 7.11015 4.8875 4.1857 3.6725 3.6730 3.1690 6.2945 5.5915 6.0945 4.5725 4.2660 4.7735 3.5755 4.8800

a T1 computed from Fs Ok Alm; T2 computed with Gt Opx exchange equilibrium; T3 computed with Gt Cd exchange equilibrium; P1 computed with Py Qz En Cd; T4 P2 computed from intersection of P1 with Alm Qz Fs fCd equilibrium. b Samples with strongly zoned garnet; P1 computed from Py Gr Qz En An. c Samples containing Grt Opx Pl Qz; P1 computed from Gr Py Qz En An d Samples containing Grt Cd Sil Pl Qz; P1 and P2 computed from Py Si Qz Cd and Gr Si Qz An, respectively; T4 T3 e Samples containing Grt Opx Pl Qz; P1 computed from Py Qz En Si; P2 computed from intersection of T2 and P1

Begin and Pattison (1994)

782 780 7.3775 6.9767

6.2750 5.9740 6.5772 6.5765 10.2800 7.8820 11.2885 10.9830 8.4842

Holtta et al. (in prep) 0.090 Aranovich 0.049 (1991) 0.091 0.113 0.079

7.8900 7.7845 9.7960 7.8970 8.4850

795 850 890 800 840

780

34

Fig. 10 P XFe diagram showing divariant loop for the assemblage Gt Cd Opx Qz. See text for discussion of evolution of two coronitic samples described by Berg (1977)

common feature in these rocks, suggesting their formation in the course of the continuous reaction: high Mg Gt Qz lower Mg Gt Opx Cd (f)

Hence the core compositions of garnets in these samples may be considered as relics of the starting material, while those of the outer parts of garnet grains should more closely correspond to equilibrium with Opx and Cd. Details are not given, however, as to which specific samples contain these textural features. For one of the samples ( 2-1572), Berg gave a detailed compositional profile of the garnet (Opx and Cd are almost homogeneous). Temperatures calculated on the basis of near-rim compositions (800 C) of minerals for this sample ( 21572r in Table 2) are much lower than those based on the core compositions (950 C) and agree very well with the estimates from other samples by Berg (1977). It can be argued that the rim compositions of garnet may not represent the original equilibrium ones because of later-stage cooling undergone by the rocks. Detailed observations show that the late-stage intergranular diffusion caused by such cooling generally affects only the very outer parts of Fe Mg minerals at their mutual contacts and is readily recognizable by the steep compositional gradient it normally produces (e.g., Berg 1977; Perchuk et al. 1985) as well as close correspondence of compositional changes in the adjacent mineral grains to the progress of the exchange reactions. Compositions of the inner parts of the grains (but still removed from the cores) which exhibit much shallower compositional gradients, or contacts with leucocratic minerals seem to be better candidates for geothermobarometry of coronabearing samples. In any event employing the core com-

positions of the minerals surrounded by coronas for retrieving P T conditions of their formation is unjustifiable and should lead to erroneous conclusions regarding the P T evolution of rocks. Conversely, in high-grade rocks lacking prograde zoning and textural evidence of reaction relationships among minerals, it seems reasonable to use core compositions for recovering near-peak physical conditions of high-grade metamorphism. Some caution is needed, however, as detailed compositional maps (Pattison and Begin, 1994) indicate that some Fe Mg reequilibration of Opx and Gt cores may occur in high temperature granulites. Formation of Opx Cd coronas is a typical feature of many high-grade metamorphic rocks of the appropriate bulk composition and is usually explained by rapid decompression (e.g. Harley 1989 and references therein). Our systematics suggest that cooling could also be a factor if primary Opx was involved in the corona-forming reaction. A significantly lower Al content of corona Opx relative to that of the matrix Opx would be characteristic in the latter case (see Fig. 7). The samples described by Berg (1977) do not, however, contain primary Opx and therefore their reaction textures should be attributed to decompression. Rocks with different bulk Fe (Fe Mg) ratio must initiate reaction during decompression at different pressures. Fig. 9 illustrates this point for several of Bergs (1977) samples. Sample 2-1572 which has a Gt core composition of Mg 30 should start to react to more Fe-rich Gt Opx Cd at about 6 kbar, while garnet with Mg 18 (sample KI3909) would remain unreacted up to 4 kbar. If core compositions of the garnets were used in P T calculations this would have produced an apparent P-difference between the two samples of about 2 kbar and the erroneous conclusion of a non-isobaric origin of the corresponding parts of the Nain complex. At the same time if the composition of the outer parts (but not the extreme rims affected by low-temperature Fe Mg exchange) of the garnet in the first sample are taken, both T and P estimates for these samples become very similar (800850 C, 44.8 kbar; Table 2). It is noteworthy that the more Fe-rich garnet (KI3909 is almost unzoned (except for the very narrow outer parts of the grains in direct contact with cordierite (Fig. 3 of Berg 1977). So far as diffusivities of Mg and Fe in garnet are similar (Chakraborty and Ganguly 1992), at least in the limited compositional range of the Fe-rich garnets, the difference in zoning profiles between the two garnets appears to be due to the difference betweenstarting (nonequilibrium core) and final (equilibrium nearrim) compositions for each of them rather than the difference in the rate of diffusional homogenization. The above example shows also how dangerous could be an attempt to calibrate garnet-orthopyroxene geothermobarometers basing on the naturally occurring assemblages, especially those containing reaction textures. Two samples (S2 and P79 in Table 2) described by Begin and Pattison (1994) contain the assemblage Gt Cd Sil Pl Qz. TWQ calculations show very good agreement between Gt-Cd exchange temperatures for

35

these samples and those that also contain Opx (B74B and B58 in Table 2), as well as with TFe Al for the latter samples. Gt-Opx exchange temperatures are considerably lower for most samples (up to 130 C), again suggesting susceptibility of Opx to compositional reequilibration at lower temperature. It should be noted, however, that the differences between our results with the two thermometers are not as extreme as Begin and Pattison (1994) calculated on the basis of Harleys (1984) experimental data. For the Aldan Shield samples containing Gt Opx Sil (last three samples in Table 2), our systematics again give TFe Al higher than those calculated from TFe Mg , suggesting some resetting of the latter thermometer for these very high temperature rocks. One of the samples (Tok-18, Aranovich and Podlesskii 1989) represents an Archean granulite basement reworked and uplifted during a major Proterozoic event in the Stanovoy fold belt, Eastern Siberia (Karsakov 1978; Perchuk et al. 1985). It is characterized by complex textural features compatible with its long P T history (Aranovich and Podlesskii 1989), and also by a very high Al2O3 content of the core of Opx (11 wt% Al2O3). This composition yields TFe Al 1000 C, which is about 170 C higher than that based on TFe Mg (Table 2). An alternative explanation to low-temperature Fe Mg reequilibration could be that garnet was formed in this rock later than Opx by the reaction (Aranovich and Podlesskii 1989): high Al, Fe Opx lower Al, Fe Opx Gt ( Sil, Qz) (g)

eter and Al2O3 content of Opx is an excellent geothermometer, independent of other compositional variables. Comparison of temperature estimates from different exchange and net-transfer reactions for a number of samples representative of high-grade terranes demonstrates the ability of carefully chosen portions of Fe Mg minerals to preserve information regarding a high temperature steady state during their evolution. Relative insensitivity of the equilibrium Fs Ok Alm to late Fe Mg reequilibration, makes this new thermometer the most robust for the Gt Opx assemblage. Applications of this thermometer yield consistent temperature estimates relative to Gt Cd Fe Mg exchange temperatures, but somewhat higher temperatures than recorded by the Gt Opx Fe Mg exchange thermometer.
Acknowledgements We gratefully acknowledge the generous support received by LYA from the Continental Geoscience Division of the Geological Survey of Canada. We also appreciate the helpful review provided by E. Froese. This paper is a contribution to IGCP 304 (Lower Crustal Processes), and is Geological Survey of Canada contribution number 41294.

References
Andersen DJ, Lindsley DH (1979) The olivine-ilmenite thermometer. Proc Loth Lunar Planet Sci Conf. Geochim Cosmochim Acta 1 (Suppl 11):493507 Andersen DJ, Bishop FC, Lindsley DH (1991) Internally consistent solution models for Fe Mg Mn Ti oxides: Fe Mg Ti oxides and olivine. Am Mineral 76: 427444 Anovitz LM (1991) Al-zoning in pyroxene and plagioclase: window on the late prograde to early retrograde P T paths in granulite terranes. Am Mineral 76: 13281343 Aranovich LY (1983) Biotite garnet equilibria in metapelites. I. Thermodynamics of solid solutions and minal reactions. In: Zharikov VV, Fedkin V (eds) Ocherki fiziko chimicheskoi petrologii. Nauka Press, Moscow, pp 1433 Aranovich LY (1991) Mineral equilibria of multicomponent solid solutions. Nauka Press, Moscow Aranovich LY, Berman RG (1995) Al2O3 solubility in orthopyroxene in equilibrium with almandine in the FeO Al2O3 SiO2 system. Current Res Geol Surv Can Pap 1995-E:271278 Aranovich LY, Berman RG (1996) A new orthopyroxene-garnet thermometer, based on reversed Al2O3 solubilities in orthopyroxene in the FeO Al2O3 SiO2 system. Am Mineral (in press) Aranovich LY, Podlesskii KK (1989) Geothermobarometry of high-grade metapelites: simultaneously operating reactions. In: Cliff RA, Yardley BWD, Daly JS (eds) Evolution of metamorphic belts. Blackwell Scientific Publications, Oxford, pp 4562 Begin NJ, Pattison DRM (1994) Metamorphic evolution of granulites in the Minto block, northen Quebec: extraction of peak P T conditions taking account of late Fe Mg exchange. J Metamorphic Geol 12: 411-428 Berg JH (1977) Regional geobarometry in the contact aureoles of the anorthositic Nain complex, Labrador. J Petrol 18: 399430 Berman RG (1988) Internally-consistent thermodynamic data for minerals in the system Na2O K2O CaO MgO FeO Fe2O3 Al2O3 SiO2 TiO2 H2O CO2. J Petrol 29: 455522 Berman RG (1991) Thermobarometry using multi-equilibrium calculations a new technique, with petrological applications. Can Mineral 29: 833855

and, therefore, the core Opx represents a relic composition which never was in equilibrium with garnet and should not be used in Gt Opx thermobarometery. TFe Al estimates consistent with TFe Mg, obtained for this sample from the composition of slightly less coarse-grained Opx (Tok-18 1; Table 2), offer support for this possibility.

Conclusions
The calculations and discussion presented in this paper demonstrate that the thermodynamic systematics derived in Part I predict phase relationships and compositions of solid solutions in good agreement with direct experimental observations that were not used in the calibration. This external consistency with a large body of experimental observations increases confidence in applications to both petrogenetic grid and thermobarometric calculations. The predicted position of the univariant FMAS equilibrium Gt Cd Opx Si Qz occurs between 7.8 kbar 700 C and 10.6 kbar 1100 C, in excellent agreement with thermobarometric results for both Opx Si Qz assemblages with the lowest Mg and Cdbearing granulites with the highest Mg . In these assemblages, garnet composition is an excellent geobarom-

36 Berman RG, Aranovich LY (1996) Optimized standard state and solution properties of minerals: I. Model calibration for olivine, orthopyroxene, cordierite, garnet, and ilmenite in the system FeO MgO CaO Al2O3 TiO2 SiO2. Contrib Mineral Petrol 126: 124 Bertrand P, Ellis DJ, Green DH (1991) The stability of sapphirinequartz and hypersthene-sillimanite-quartz assemblages: an experimental investigation in the system FeO MgO Al2O3 SiO2 under H2O and CO2 conditions. Contrib Mineral Petrol 108: 5571 Bishop FC (1980) The distribution of Fe2 and Mg between coexisting ilmenite and pyroxene with application to geothermometry. Am J Sci 280: 4677 Bohlen SR, Boettcher AL (1981) Experimental investigations and geological applications of orthopyroxene geobarometry. Am Mineral 66: 951964 Bohlen SR, Essene EJ, Boettcher AL (1980) Reinvestigation and application of olivine-orthopyroxene-quartz barometry. Earth Planet Sci Lett 47: 110 Bohlen SR, Wall VJ, Boettcher AL (1983) Experimental investigations and geological applications of equilibria in the system FeO TiO2 Al2O3 SiO2 H2O. Am Mineral 68: 10491058 Chakraborty S, Ganguly J (1992) Cation diffusion in aluminosilicate garnets: experimental determination in spessartine-almandine diffusion couples, evaluation of effective binary diffusion coefficients, and applications. Contrib Mineral Petrol 111: 7486 Coolen JJM (1980) Chemical petrology of the Furua granulite complex, southern Tanzania. GUA Pap Geol 13: 1258 de Capitani C, Brown TH (1987) The computation of chemical equilibrium in complex systems containing non-ideal solutions. Geochim Cosmochim Acta 51: 26392652 Eckert JO, Bohlen SR (1992) Reversed experimental determinations of the Mg Fe2 exchange equilibrium in Fe-rich garnetorthopyroxene pairs. EOS Trans Am Geophys Union 73: 608 Ellis DJ, Sheraton JW, England RN, Dallwitz WB (1980) Osumilite-saphhirine-quartz granulites from Enderby Land, Antarctica-mineral assemblages and reactions. Contrib Mineral Petrol 72: 123143 Finnerty AA, Boyd FR (1984) Evaluation of thermobarometers for garnet peridotites. Geochim Cosmochim Acta 48: 1527 Fitzsimons ICW, Harley SL (1994) Disequilibrium during retrograde cation exchange and recovery of peak metamorphic temperatures: a study of granulites from Antarctica. J Petrol 35: 543576 Florence FP, Spear FS (1991) Effects of diffusional modification of garnet growth zoning on P T path calculations. Contrib Mineral Petrol 107: 487500 Fonarev VI (1987) Mineral equilibria in Precambrian iron formations. Nauka Press, Moscow Fonarev VI, Korolkov GJ, Dokina TN (1976) Stability of the orthopyroxene magnetite quartz association under hydrothermal conditions. Geochem Int 13: 134146 Frost BR, Chacko T (1989) The granulite uncertainty principle: limitations on thermobarometry in granulites. J Geol 97: 435450 Green DH, Sobolev NV (1975) Coexisting garnets and ilmenites synthesized at high pressures from pyrolite and olivine basanite and their significance for kimberlitic assemblages. Contrib Mineral Petrol 50: 217229 Harley SL (1984) The solubility of alumina in orthopyroxene coexisting with garnet in FeO MgO AL2O3 SiO2 and CaO FeO MgO Al2O3 SiO2. J Petrol 25: 665696 Harley SL (1989) The origins of granulites: a metamorphic perspective. Geol Mag 126: 215 247 Harlov DE (1992) Comparative oxygen barometry in granulites, Bamble Sector, SE Norway. J Geol 100: 447464 Hayob JL, Bohlen SR, Essene EJ (1993) Experimental investigation and application of the equilibrium rutile orthopyroxene quartz ilmenite. Contrib Mineral Petrol 115: 1835 Helgeson HC, Delany JM, Nesbitt HW, Bird DK (1978) Summary and critique of the thermodynamic properties of rock-forming minerals. Am J Sci 278A Hensen BJ (1976) The ability of pyrope-grossular garnet with excess silica. Contrib Mineral Petrol 55: 279292 Hensen BJ, Harley SL (1990) Graphical analysis of P T X relations in granulite facies metapelites. In: Ashworth JR, Brown M (eds) High-temperature metamorphism and crustal anatexis. Unwin Hyman, United Kingdom, pp 1956 Holland TJB, Powell RP (1990) An internally consistent thermodynamic dataset with uncertainties and correlations: the system K2O Na2O CaO MgOMnOFeOFe2O3 Al2O3 TiO2 SiO2 C H2 O2. J Metamorphic Geol 8: 89124 Jaffe HW, Robinson P, Tracy RJ (1978) Orthoferrosilite and other iron-rich pyroxenes in microperthite gneiss of the Mount Marcy area, Adirondack Mountains. Am Mineral 63: 1116 1136 Jen LS, Kretz R (1981) Mineral chemistry of some mafic granulites from the Adirondack region. Can Mineral 19: 479491 Johnson CA, Essene EJ (1982) The formation of garnet in olivinebearing metagabbros from the Adirondacks. Contrib Mineral Petrol 81: 240251 Karpov IK, Kiselev AI, Letnikov FA (1976) Computer modeling of natural mineral formation processes (in Russian). Nauka Press, Moscow Karsakov LP (1978) Deep-seated granulites (in Russian). Nauka Press, Moscow Kawasaki T, Matsui Y (1983) Thermodynamic analyses of equilibria involving olivine, orthopyroxene and garnet. Geochim Cosmochim Acta 47: 16611679 Koch-Muller M, Cemic L, Langer K (1992) Experimental and thermodynamic study of Fe Mg exchange between olivine and orthopyroxene in the system MgO FeO SiO2. Eur J Mineral 4: 115135 Korzhinskii DS (1973) Theoretical basis for analysis of parageneses of minerals (in Russian). Nauka Press, Moscow Koziol AM, Bohlen SR (1992) Solution properties of almandinepyrope garnet as determined by phase equilibrium experiments. Am Mineral 77: 765773 Lamb RC, Smalley PC, Field D (1985) P T conditions for the Arendal granulites, southern Norway: implications for the roles of P, T and CO2 in deep crustal LILE-depletion. J Metamorphic Geol 4: 143160 Lee HY, Ganguly J (1988) Equilibrium compositions of coexisting garnet and orthopyroxene: experimental determinations in the system FeO MgO Al2O3 SiO2, and applications. J Petrol 29: 93113 Lindsley DH (1980) Phase equilibria of pyroxenes at pressures 1 atmosphere. In: Prewitt CT (ed) Pyroxenes (Reviews in Mineralogy vol 7). Mineralogical Society of America, Washington DC, pp 289308 Matsui Y, Nishizawa O (1974) Iron(II)-magnesium exchange equilibrium between olivine and calcium-free pyroxene over a temperature range 800 C to 1300 C. Bull Soc Fr Mineral Cristallogr 97: 122130 Medaris Jr LG (1969) Partitioning of Fe and Mg between coexisting synthetic olivine and orthopyroxene. Am J Sci 267: 945 968 Mora CI, Valley JW (1985) Ternary feldspar thermometry in granulites from the Oaxacan Complex, Mexico. Contrib Mineral Petrol 89: 215225 Newton RC (1972) An experimental determination of the highpressure stability limits of magnesian cordierite under wet and dry conditions. J Geophys Res 80: 398420 Pattison DRM, Begin NJ (1994) Compositional maps of metamorphic orthopyroxene and garnet: evidence for a hierarchy of closure temperatures and implications for geothermometry of granulites. J Geol 12: 387410 Pattison DRM, Newton RC (1989) Reversed experimental calibration of the garnet-clinopyroxene Fe Mg exchange thermometer. Contrib Mineral Petrol 101: 87103 Perchuk LL, Aranovich LY, Podlesskii KK, Lavrenteva IV, Gerasimov VY, Kitsul VI, Korsakov LP, Berdnikov NV (1985) Precambrian granulites of the Aldan shield, eastern Siberia, the USSR. J Metamorphic Geol 3: 265310

37 Percival JA (1991) Granulite-facies metamorphism and crustal magmatism in the Ashuanipi Complex, Quebec Labrador, Canada. J Petrol 32: 12611297 Perkins D (1983) The stability of Mg-rich garnet in the system CaO MgO Al2O3 SiO2 at 1000 1300 C and high pressure. Am Mineral 68: 355364 Perkins DI, Essene EJ, Marcotty LA (1982) Thermometry and barometry of some amphibolite-granulite facies rocks from the Otter Lake area, southern Quebec. Can J Earth Sci 19: 17591774 Powell R, Holland TJB (1985) An internally consistent thermodynamic dataset with uncertainties and correlations: 1. Methods and a worked example. J Metamorphic Geol 3: 327342 Powell R, Holland TJB (1988) An internally consistent thermodynamic dataset with uncertainties and correlations: 3. Applications to geobarometry, worked examples and a computer program. J Metamorphic Geol 6: 173204 Prigogine I, Defay R (1954) Chemical thermodynamics. Longmans, Norwich Sack RO, Ghiorso MS (1989) Importance of considerations of mixing properties in establishing an internally consistent thermodynamic database: thermochemistry of minerals in the system Mg2SiO4 Fe2SiO4 SiO2. Contrib Mineral Petrol 102: 4168 Sandiford M (1985) The metamorphic evolution of granulites at Fyfe Hills: implications for Archaean crustal thickness in Enderby Land, Antarctica. J Metamorphic Geol 3: 155178 von Seckendorff V, ONeill HSC (1993) An experimental study of Fe Mg partitioning between olivine and orthopyroxene at 1173, 1273 and 1473 K and 1.6 Gpa. Contrib Mineral Petrol 113: 196207 Wood BJ, Holloway JR (1984) A thermodynamic model for subsolidus equilibria in the system CaO MgO Al2O3 SiO2. Geochim Cosmochim Acta 48: 159176